Recent progress in parallel stochastic approximations
Parallel recursive algorithms of stochastic approximation type are discussed. The basic idea is to utilize a collection of parallel processors in lieu of using a single one alone as in the classical algorithms. The processors compute and communicate with each other asynchronously. By asynchronous implementation we mean that each processor updates on its own pace, and uses the most recently recieved information from other processors. Several procedures are dealt with. These include distributed communication schemes via convexifications, parallel algorithms with interacting processors and renewal type of computation times, and real time implementable procedures with pipeline arrangements. It appears that from the close connection between real world applications and mathematical theories, the study of parallel stochastic approximation methods gains not only much of its special appeal, but also much of its intrinsic tension. The main emphasis and interest are the asymptotic properties of the algorithms. The parallel interacting algorithm and its variation are used as representatives for illustrating various limiting results. Convergence and rate of convergence issues are addressed, and robustness analysis is carried out. For the entire collection see [Zbl 0778.00024].
- Stochastic approximation algorithms for parallel and distributed processing
- On W.P.1 Convergence of A Parallel Stochastic Approximation Algorithm
- Parallel and bootstrapped stochastic approximation
- Asymptotic Properties of Distributed and Communicating Stochastic Approximation Algorithms
- On robustness of the Robbins-Monro method for parallel processing
- On robustness of the Robbins-Monro method for parallel processing
- Joint robustness on noise and Lyapunov functions for parallel stochastic approximation algorithms
- On W.P.1 Convergence of A Parallel Stochastic Approximation Algorithm
- scientific article; zbMATH DE number 4157849 (Why is no real title available?)
- Asymptotic Properties of Distributed and Communicating Stochastic Approximation Algorithms
- Stochastic approximation algorithms for parallel and distributed processing
- scientific article; zbMATH DE number 550187 (Why is no real title available?)
- scientific article; zbMATH DE number 1163645 (Why is no real title available?)
- Parallel and bootstrapped stochastic approximation
- Optimal quasi-convex combinations for stochastic approximation algorithms with parallel observers
- On the rates of convergence of parallelized averaged stochastic gradient algorithms
- Stochastic approximation
This page was built for publication: Recent progress in parallel stochastic approximations
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3140597)